Meta says it is granting $2 million across five US universities for research that uses its AI models, from a camera system that watches over infants with breathing tubes to tools that could flag heart failure earlier. The company announced the programme on 28 October.
The universities are the California Institute of Technology (Caltech), The Ohio State University College of Medicine, Louisiana State University (LSU), Indiana University and UTHealth Houston. As well as money, Meta is offering access to its models and direct collaboration with its researchers. A Meta research blog post says support “varies by institution and project”, with funding for selected projects; Meta has not said how the $2 million is split.
Which models the researchers will use
Researchers will build with Meta’s Muse family, including Muse Spark Open, which Meta calls its most capable open-weight model and whose weights it released this month, and the Muse Wonder and Muse Spark series. Researchers also get Muse Glimmer, a 30-billion-parameter open-weight model that Meta says runs on a single consumer graphics card. Projects also use Meta’s computer vision models SAM, V-JEPA and DINO, its SeamlessM4T translation model and, in some cases, its glasses and headsets.
Meta says the faculty brought their own research questions, designed their studies and decided how to use the tools. For the healthcare projects, the universities will run Meta’s models in their own environments and keep control of all patient data, which Meta says will not be shared with or processed by the company.
The medical projects
At LSU, Dr Michael Dunham leads Eyes-On, a computer vision system meant to detect when a tracheostomy tube slips out of place in an infant, a life-threatening emergency in babies who cannot cry out for help. It is exploring SAM for tracking and is designed for bedside cameras. Meta says a proof of concept has been published. Other LSU work includes a neurosurgical outcomes registry, AI coaching for medical training in virtual reality, and models for head and neck cancer.
At Ohio State, Dr Yuan Xue, cardiologist Dr Yuchi Han and Dr Xia Ning want to predict heart failure with preserved ejection fraction, which accounts for roughly half of heart failure cases, from routine abdominal CT scans, medical records and ECG data. The idea of reusing data already collected echoes recent work on spotting cancer risk in routine records. Ohio State teams will also draft chest X-ray reports with AI and build a clinician-led agent system for Alzheimer’s diagnosis.
At Indiana, Dr David Crandall is studying how conversational AI, including social robots, could adapt to people living with dementia. Another Indiana project, NeoSmartML+, aims to give offline, multilingual guidance on newborn skin-to-skin care in low- and middle-income countries. UTHealth Houston is testing Muse Glimmer on its own secure hardware for clinical agents, including work on glioblastoma records.
Science beyond medicine
Caltech’s expanded partnership tests whether large language models can do original research rather than summarise it. Projects include MathDB Frontier, a benchmark built on more than 86,900 research-level open maths problems, a chemistry test using unpublished lab notebook records, and TorchLean, a framework for producing formally checkable results.
Open weights make hospital AI easier to keep in-house
The detail that matters most for hospitals is where the models run. Patient records are highly sensitive, and running a model inside the hospital avoids sending them to an outside company’s servers. Because Meta’s open-weight models can be downloaded, the universities can run them on their own machines and, according to Meta, keep patient data away from the company. That is a practical advantage open models hold over services reached only through the cloud.
The sums are small. Meta has not said how the $2 million is split, and spread across five institutions and more than a dozen projects it will not go far. For Meta, the bigger value may be researchers proving what its Muse models can do on hard scientific and clinical problems, much as Anthropic has done by publishing Claude’s drug-development results.
Most of the medical projects are at an early stage. Only the LSU infant-monitoring work has a published proof of concept, and UTHealth Houston describes its glioblastoma tests on 65 patient cases as workflow checks rather than clinical outcomes. Whether any of this improves patient care will depend on results that have not yet been reported.
What happens next
Meta says it will share more as the projects get under way. Some teams expect to publish more: LSU’s virtual reality coaching group plans to share user studies, and its head and neck cancer team has manuscripts in preparation.
Sources
- Meta Newsroom, “Meta Partners With Five US Universities to Accelerate Medical and Scientific Breakthroughs”, 28 October 2026
- Meta AI Research, “How American Universities Are Using AI Models From Meta to Accelerate Medical and Scientific Breakthroughs”, 28 October 2026
- AI at Meta on X, 28 October 2026
Related stories
- Meta Releases the Weights for Muse Spark Open, and Apple Shows It Running Across Two to Four Mac Studios
- Meta Launches Muse Wonder and Muse Wonder Full, Its “Most Capable Models to Date”, Built for AI Agents
- ESMO Study: Routine Health Data Flags Higher Risk of 20 Cancers
- Anthropic Reports Claude Drug Work at Six Partners and Buys Mithril
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
